Paranemin and the organization of desmin filament networks

Intermediate filaments (IFs) can be formed from a large number of related subunit proteins. Based on sequence, gene structure and assembly properties, cytoplasmic IF proteins have historically been divided into distinct groups: the keratins (types I and II), the vimentin-like proteins (type III), and the neurofilament-like proteins (type IV) (Quinlan et al., 1994). Recently, the existence of an additional group of very large IF proteins has been recognized and variously classified as belonging to the neurofilament-like group (Herrmann and Aebi, 2000) or constituting a distinct class (type VI) (Steinert et al., 1999).

IF proteins share a canonical three-domain structure consisting of a structurally conserved central α-helical ‘rod’ domain flanked by more variable nonhelical N-terminal ‘head’ and C-terminal ‘tail’ domains (Herrmann and Aebi, 2000). The effects of various mutations on the in vitro and in vivo assembly properties of IF proteins have shown that specific sequences within the head and rod domains are strictly required for IF assembly (see Heins and Aebi, 1994). The tail domain is not required for IF assembly, but appears to affect the diameter of assembled filaments (Rogers et al., 1995). Although IF proteins share common structural features, there is considerable variability in primary sequence, particularly in the nonhelical terminal domains, where there is limited or no direct sequence similarity between IF-protein groups (Quinlan et al., 1994).

IF proteins exhibit characteristic differences in their assembly properties that can affect the subunit composition of IF networks, depending on the IF-proteins that are expressed by a given cell type. The keratins, neurofilament-like proteins and large IF proteins are obligate heteropolymers. The vimentin-like proteins vimentin, desmin, glial fibrillary acidic protein (GFAP) and peripherin will coassemble with each other, the neurofilament proteins and the large IF proteins, but not with the keratins. In contrast to the other IF-proteins, each of the vimentin-like proteins can also form homopolymeric filaments (Herrmann and Aebi, 2000). Although the vimentin-like proteins can form homopolymeric IFs, in some cell types they coassemble with large IF proteins to form heteropolymeric filament networks (Herrmann and Aebi, 2000). Nestin (Lendahl et al., 1990), synemin (Bellin et al., 1999) and paranemin (Hemken et al., 1997) are large IF proteins that are related in the sense that they consist of very short, nonhelical, N-terminal ‘head’ domains, a central α-helical ‘rod’ domain characteristic of IF-proteins, and unique, very long C-terminal ‘tail’ domains that account for most of the high molecular mass of these proteins (Herrmann and Aebi, 2000). Nestin is transiently expressed with vimentin in cells of the developing central nervous system (Tohyama et al., 1992; Dahlstrand et al., 1995) and myoblasts (Sejersen and Lendahl, 1993). Synemin and paranemin are expressed with desmin in muscle cells (Hemken et al., 1997; Bilak et al., 1998; Carlsson et al., 2000). Synemin is also transiently expressed in the developing central nervous system, although the pattern of expression differs from that of nestin (Sultana et al., 2000). The role of these large IF-proteins in IF organization is not well understood. Eliasson et al. have proposed that nestin may be required for vimentin IF assembly in immature glial cells (Eliasson et al., 1999), whereas it has also been suggested that synemin (Bellin et al., 1999) and paranemin (S. A. Seiler and R. M. Robson, unpublished observations) may crosslink desmin IFs to other cytoskeletal structures. However, there has been no direct analysis of the function of these large IF proteins.

There is growing evidence that the subunit composition of IFs can have dramatic effects on how filaments are organized in particular cell types. Myoblasts initially express vimentin and nestin (Sejersen and Lendahl, 1993). Desmin synthesis is induced early in myogenic differentiation, and desmin is integrated into the preexisting vimentin filament network (Bennett et al., 1979; Gard et al., 1979; Capetanaki et al., 1997). In the chick, paranemin and synemin expression are also induced early in myogenesis (Price and Lazarides, 1983). The level of vimentin (Bennett et al., 1979) then declines and desmin assembled with large IF proteins comprise the IFs characteristic of mature muscle. There appears to be some species variation in the expression of the large IF proteins during myogenesis. In humans, nestin expression declines and is not detectable in mature muscle (Kachinsky et al., 1994; Sjoberg et al., 1994), while in rats nestin is downregulated but remains in detectable quantities (Sejersen and Lendahl, 1993; Carlsson et al., 1999). Paranemin and synemin were originally characterized as components of chicken muscle (Price and Lazarides, 1983). Subsequent studies have shown that both synemin (Bilak et al., 1998) and paranemin (Hemken et al., 1997) are expressed in porcine muscle. Recently, Carlsson et al. reported that nestin (Carlsson et al., 1999) as well as paranemin and synemin (Carlsson et al., 2000) are detectable in mature mouse muscle. This change in IF protein expression is concomitant with a fundamental change in the organization of the IF network in these cells. The longitudinal IF network characteristic of early myotubes is reorganized into a transverse network that associates with the Z-lines, sarcolemma and elements of the myotendenous junction (Tokuyasu et al., 1984; Tokuyasu et al., 1985). In Xenopus laevis, myotomal muscle precursors do not express vimentin, and the desmin network forms de novo (Cary and Klymkowsky, 1994a). Desmin is found associated with the sarcolemma and never forms a longitudinal network, whereas expression of exogenous vimentin leads to the formation of a longitudinal network (Cary and Klymkowsky, 1994b).

To determine if the differences in primary structure among the vimentin-like proteins have some characteristic effect on the organizational behavior of IFs, we conducted studies to compare vimentin, desmin, GFAP and peripherin IFs assembled under identical conditions in IF-free SW-13 cells. Whereas vimentin, GFAP and peripherin form extended IF networks de novo, desmin network organization was qualitatively different. The formation of extended desmin filament networks required the additional presence of either vimentin, paranemin or a keratin IF system. Although synemin and nestin colocalized with desmin, they did not support the formation of an extended IF network. These studies indicate that desmin has different organizational properties from the other vimentin-like IF proteins, and implicate interaction with paranemin as particularly important in the organization of desmin IF networks.

All cells were maintained in F12:DMEM (1:1) containing 5% fetal bovine serum. SW-13/cl.1 vim⁺ and SW-13/cl.2 IF-free cell lines were grown in monolayer culture as previously described (Sarria et al., 1994). SW-13/cl.2 cells that express human keratins K8 and K18 (T7K) were provided by K. Trevor (University of Arizona Cancer Center) and SW-13/cl.2 cells that express peripherin were provided by R. Liem (Columbia University). Transfected cells were maintained in medium containing 200 μg/ml G418. MFT6 and MFT16 cell lines were derived from vim^+/+ and vimentin^−/− primary embryo fibroblasts, respectively, by stable transfection with SV-40 genes (Holwell et al., 1997). MCF-7 cells were cultured as described previously (Sarria et al., 1990).

pRC/CMVdesmin, encoding mouse desmin (Li et al., 1994) was used without modification. To obtain a vimentin expression plasmid with a similar CMV promoter element, the mouse vimentin cDNA was released from pSP64-MMTV-Vims (Sarria et al., 1990) by digestion with BamH1, and inserted into the BamH1 site of pcDNA 3, resulting in pcDNA3-mVim. An analogous GFAP expression plasmid was produced by digestion of pGEM:GFAP (gift of R. Liem, Columbia University) with EcoR1 and insertion of the recovered rat GFAP cDNA into the EcoR1 site of pcDNA 3, resulting in pcDNA 3/GFAP. To construct a mouse vimentin head–desmin rod and tail chimera, the head domain for vimentin (5′ oligo: CCCCATatgtctaccaggtctgtg and 3′ oligo: CCCGTCgaccgagtcttgaagc) and the body and tail domains of desmin (5′ oligo CCCCTCgagttctccctggccgacg and 3′ oligo: CCCTCTAGAcagcacttcatgttgttg) were amplifed by PCR. Amplified vimentin head DNA was digested with NdeI and SalI, while desmin DNA was digested with XhoI and XbaI (sites underlined) restriction enzymes, and subcloned into the pT7-myc plasmid, digested with NdeI and XbaI, to form pT7-mVDchimeria-myc (pT7mVD-myc). A single, conservative D-to-E change occurs in the junction between the vimentin and desmin-derived regions. The cDNA encoding mVDmyc was released from pT7m VDchimera-myc by digestion with HindIII and BamHI and subcloned into the HindIII/BamHI sites of pcDNA3 to make pcDNA-mVDmyc. To construct a mouse desmin head–vimentin rod and tail chimera, the head domain for desmin (5′ oligo: GGAAGCTTatgagccaggcctactcgtccagc and 3′ oligo: GGGTCgacgagcaaccacgcgcaccggtcct) and the body and tail domains of vimentin (5′ oligo GGCTCGAGaagaacacccgcaccaacgagaag and 3′ oligo: GGCCTAGGttattccagtagcactacgactct) were amplifed by PCR. Amplified desmin head DNA was digested with HindIII and SalI, while vimentin DNA was digested with XhoI and BamHI, and then subcloned into the HindIII/BamHI sites of pcDNA3 to make pcDNA-DV. Expression plasmids encoding rat nestin (pNg), HA-epitope tagged rat nestin (pNG-HA) (gifts of M. Marvin, Harvard Medical School) and avian paranemin (pRC/RSV-paranemin) and synemin (pcDNA3-synemin) have been previously described (Marvin et al., 1998; Hemken et al., 1997; Bellin et al., 1999) and were used without modification.

To produce stable cell lines expressing desmin, vimentin or GFAP, approximately 1.5×10⁶ SW-13/cl.2 cells were plated in a 10 cm culture dish and transfected with 40 μg of pRC/CMVdesmin, pcDNA3/mVim, or pcDNA3/GFAP plasmid DNA using calcium phosphate as previously described (Sarria et al., 1994). Stable transfectants were selected in medium containing 400 μg/ml G418, and isolated G418-resistant colonies were screened using anti-IF immunofluorescence. Stable lines coexpressing desmin and keratins K8 and K18 were obtained by transfection of SW-13/T7K cells with 40 μg pRC/CMV desmin and 4 μg pCMVHygro using calcium phosphate. Stable transfectants were selected in medium containing 100 μg/ml G418 and 100 μg/ml hygromycin. Cells were transiently transfected in 60 mm culture dishes with 20 μg DNA either by calcium phosphate precipitation (Graham and van der Eb, 1973), or by a modification of this protocol using Calphos Maximizer (Clonetech) as indicated by the manufacturer. The cells were incubated with the Ca-DNA solution for 18 hours, the monolayer was rinsed with growth medium, and the cells incubated for 24 or 48 hours before being processed for immunofluorescence microscopy.

Cells were plated on sterile glass coverslips. The coverslips were rinsed in PBS, and fixed with 70% acetone, 30% methanol for 10 minutes at −20°C. To visualize IF proteins, indirect immunofluorescence was performed with primary rabbit anti-desmin (gift of G. Miller, University of Colorado Health Sciences Center), monoclonal anti-desmin antibody (Dako), human vimentin-specific monoclonal anti-vimentin (clone V-9, Roche Molecular Biochemicals), a rodent vimentin-specific rabbit anti-vimentin antiserum (Sarria et al., 1990), monoclonal anti-keratin 18 antibody (Sigma), monoclonal anti-GFAP (Sigma), monoclonal anti-nestin (Sigma), rabbit anti-peripherin (gift of R. Liem, Columbia University), rabbit anti-paranemin (Hemken et al., 1997), and rabbit anti-synemin (Bellin et al., 1999). HA epitope-tagged nestin was detected using a monoclonal anti-HA (Roche Molecular Biochemicals). Alexa 488 (Molecular Probes), fluorescein or rhodamine-conjugated affinity-purified anti-mouse Ig or rhodamine-conjugated-anti-rabbit Ig (Roche Molecular Biochemicals) were used as second antibodies. All antibodies were diluted in PBS containing 1% ovalbumin and 1% normal goat serum. The coverslips were mounted in immunomount (Lerner Laboratories) and viewed with an Olympus microscope equipped with epifluorescence optics. Photomicrographs were made with Kodak TMAX 400 film and developed with HC-110 developer (Kodak) at an approximate exposure index of 1100.

Cells were labeled with 10-50 μCi/ml [³⁵S]-methionine for 4 hours in methionine-free medium containing 1% dialyzed fetal bovine serum. Triton-insoluble cytoskeletons were prepared as previously described (Evans, 1984). Samples of Triton-insoluble protein were normalized for total ³⁵S dpm, and analyzed by two-dimensional gel electrophoresis (2D-PAGE) (O’Farrell, 1975). Following electrophoresis, gels were stained with Coomassie Blue, destained, and infiltrated with 1 M sodium salicycilate. The gels were dried and fluorographed on Hyperfilm (Amersham) at −70°C.

Cells in monolayer culture were fixed in 3% buffered glutaraldehyde and then postfixed in 1% phosphate-buffered osmium tetroxide. The monolayers were then dehydrated in a graded ethanol series and embedded in a mixture of Epon and Araldite (Mollenhauer, 1992). Thin sections were examined using a Philips 201 electron microscope.

SW-13/cl.2 cells do not have cytoplasmic IFs (Sarria et al., 1990) and are therefore very useful for studying de novo IF network assembly. To obtain SW-13/cl.2 cells that express desmin, the cells were stably transfected with a mouse desmin cDNA. Immunofluorescence microscopy of independently isolated cell lines (TmD cells) revealed that in most of the cells the exogenous desmin formed short, thickened filamentous structures and prominent spot-like desmin aggregates, which were usually located in a perinuclear region (Fig. 1B,D). Although individual TmD cell lines differed in the amount of desmin that was expressed, reflected in the size of the aggregates and the presence of additional peripheral spot-like accumulations, the overall organization of desmin was similar. Extended filament networks were not observed. The inability of desmin homopolymers to form extended IF networks is not unique to SW-13 cells. Transient transfection of desmin into MFT-16 cells, a mouse fibroblast cell line derived from vimentin-null mice (Holwell et al., 1997), resulted in the formation of similar short, thickened, filamentous structures and prominent spot-like desmin aggregates that were similar to those observed in SW-13 cells (data not shown). About 1% of the parental SW-13/cl.2 cells re-express vimentin (Sarria et al., 1990). In cells in which vimentin was re-expressed, desmin formed an extended filament network (Fig. 1C). Similar desmin-vimentin networks were observed when desmin was expressed in human SW-13/cl.1 cells and mouse MFT-6 fibroblasts that contain vimentin IFs. In both cases exogenous desmin was incorporated into the extended endogenous vimentin IF network (data not shown; see Quinlan and Franke, 1982).

To determine whether desmin expressed in SW-13/cl.2 cells assembled into filaments or non-filamentous aggregates, the desmin-transfected cell line TmD 9 was examined using thin-section electron microscopy (Fig. 2). TmD 9 cells contained prominent perinuclear structures that are not observed in untransfected cells. These perinuclear structures were composed of densely packed accumulations of 9-11 nm diameter filaments (Fig. 2). These results indicate that although desmin is competent to assemble into IFs, the filaments are not able to assume an extended network.

Vimentin (Sarria et al., 1990), GFAP (Chen and Liem, 1994) and peripherin (Cui et al., 1995; Ho et al., 1995) assemble to form extended homopolymeric filament networks when expressed in SW-13/cl.2 cells. To determine whether the inability of desmin to form extended filaments was due to a significant difference between these proteins or to an idiosyncrasy in our cell line or culture conditions (for example, see Tolle et al., 1987), SW-13/cl.2 cells were transfected with either vimentin or GFAP cDNAs, and stable cell lines that express vimentin and GFAP were isolated. In addition, SW-13/cl.2 cells that express a peripherin transgene (Ho et al., 1995) were obtained for comparison under identical culture conditions. As shown in Fig. 3A, SW-13/cl.2 cells transfected with a mouse vimentin transgene regulated by the same promoter element used in the desmin constructs, contained extended filament networks. Similar extended filaments were formed in cell lines that expressed either peripherin (Fig. 3B), or GFAP (Fig. 3C). Based on these results, we concluded that the organizational properties of desmin are different from those of the other vimentin-like proteins.

To determine if the formation of desmin filament aggregates in IF-free cells might be due to differences in the amount of expression, the levels of desmin expression in the TmD cell lines shown in Fig. 1 were examined by 2D-PAGE analysis of Triton-insoluble proteins. Although TmD 9 and TmD 42 cells contained significant amounts of desmin, relative to other proteins such as lamin B (Fig. 4), neither cell line contained extended desmin filaments (Fig. 1). The level of desmin expression in the TmD 9 cells appeared to be the highest of the TmD cell lines examined, but was not obviously greater than vimentin expression in an SW-13/cl.2 derived cell line that expresses a vimentin transgene (data not shown). Consistent with the immunofluorescence microscopy, the TmD 36 cells contained significant amounts of vimentin, at clearly a much higher level of expression than the desmin transgene (Fig. 4B). The isoelectric focusing patterns of the expressed desmin transgene in the 2D-PAGE did not indicate any differences in the relative amounts of the more acidic forms of the protein, and suggested that desmin was not hyperphosphorylated in the TmD cell lines. Finally, the detergent insolubility of the desmin expressed in the transfectant cell lines (data not shown) is consistent with the observation that the desmin aggregates in these cells are assembled homopolymers.

MCF-7 is a breast cancer epithelial cell line that contains keratin IFs but not vimentin (Glass and Fuchs, 1988; Sarria et al., 1990). Interestingly, Raats et al. had found that desmin forms extended IF networks when expressed in this cell line (Raats et al., 1990; Raats et al., 1992). To determine if a keratin IF network was sufficient to rescue the network forming ability of desmin, two experiments were carried out. First, MCF-7 cells were transiently transfected with a mouse desmin expression plasmid. Desmin clearly formed a distinct and extended IF network in some but not all of the desmin-positive cells. In agreement with previous studies (Raats et al., 1990), desmin IFs were frequently found to codistribute with the keratin IF network (Fig. 5).

Next, we examined the organization of desmin in the SW-13/cl.2 derived cell line T7K, which stably expresses keratins K8 and K18. In both transient (not shown) and stable transfectants (Fig. 6), keratin and desmin formed distinct networks. The presence of a keratin IF network influenced the ability of desmin to form extended filament networks and was related to the relative levels of K8/K18 expression. In cells that contained higher levels of keratin, compared with desmin, desmin was able to form extended filament networks (Fig. 6D). In cells that contained high levels of desmin compared to keratin, desmin was present in aggregates and a partial collapse of the keratin network was observed that colocalized with the desmin aggregate (Fig. 6F). The high level of keratin expressed in MCF-7 cells (Sarria et al., 1990) presumably explains why desmin was organized into extended IF networks in these cells but not in the IF-free cells.

In muscle cells desmin forms heteropolymeric filaments with paranemin, synemin and nestin that are present at a low stoichiometry (Herrmann and Aebi, 2000). Nestin is more transiently expressed during myogenesis (Sejersen and Lendahl, 1993), whereas paranemin and synemin are IF components of developing and mature muscle (Hemken et al., 1997; Bilak et al., 1998; Carlsson et al., 2000). None of these proteins are detectable in IF-free SW-13/cl.2 cells (Hemken et al., 1997; Bellin et al., 1999; Marvin et al., 1998). To test the hypothesis that the impaired capacity of desmin to form extended filament networks in these cells might be due to the absence of large IF proteins, we transiently expressed paranemin, synemin and nestin in SW-13 cells. In control experiments, and in agreement with previous studies (Hemken et al., 1997; Bellin et al., 1999; Marvin et al., 1998), none of these large IF proteins were capable of forming filament networks on their own when expressed in IF-free SW-13/cl.2 cells (data not shown). All three proteins codistributed with vimentin when expressed in an SW-13 cell line that contained mouse vimentin (data not shown). When expressed in SW-13 derived TmD-9 cells, both synemin and nestin clearly codistributed with the desmin aggregates but had no obvious effect on desmin organization, even when expressed at apparently high levels (Fig. 7A-D). The small percentage of cells expressing either nestin or synemin that exhibited extended desmin networks was no different than that found in mock-transfected TmD 9 cells (Table 1), and is about the percentage of TmD 9 cells that express vimentin, suggesting that this represents coassembly with vimentin. In contrast, expression of paranemin, which also codistributed with desmin, induced a dramatic change in desmin organization (Fig. 7E,F). Approximately half of the TmD-9 cells that expressed detectable paranemin had extended paranemin-desmin IF networks (Table 1). On the basis of the intensity of the paranemin fluorescence, there was a relationship between the level of paranemin expression and the degree to which desmin filaments formed extended networks. Even cells that expressed a significant level of paranemin often contained smaller residual spot-like desmin IF aggregates in addition to the extended IF network (see Fig. 7E,F).

Desmin and vimentin are highly related proteins. They differ most dramatically in sequence identity in their non-helical head domains (Quinlan et al., 1994). When head domains of Xenopus desmin and vimentin were ‘swapped’ with one another and expressed in Xenopus embryos, the organization of the chimeric IF protein in the dorsal myotome took on the organizational characteristics of the filament type of the head domain (Cary and Klymkowsky, 1994a). Using this approach, SW-13/cl.2 cells were transiently transfected with epitope tagged head/body domain-swapped chimeric mouse cDNAs encoding either the desmin head domain with the body of vimentin (DV) or the vimentin head domain with the body of desmin (VD), and the organization of the chimeric homopolymers was examined. Expression of the DV chimeric protein resulted in the formation of anti-myc reactive structures that were similar to the wild-type vimentin (Fig. 8). In contrast, expression of the VD chimeric protein resulted in the formation of punctate or spot-like aggregates that were not observably different from the filaments formed by the wild-type desmin. These results indicate that, unlike the situation in Xenopus embryo, interactions involving the non-helical amino-terminal ‘head’ domain are not responsible for the differences in the organization of desmin and vimentin.

During the development of muscle in mammals, five IF proteins, namely vimentin, desmin, nestin, synemin and paranemin, are expressed at various times. This complicates efforts to sort out the contribution of individual IF proteins to the IF network. Although desmin and the related IF proteins vimentin, GFAP and peripherin coassemble in vitro and in vivo, there is growing evidence that these IF proteins are not identical in their ‘networking’ abilities (Cary and Klymkowsky, 1994a). The difference in the network-forming ability of desmin and the other vimentin-like proteins can be seen quite dramatically in IF-free SW13/cl.2 cells. While vimentin, GFAP and peripherin assemble into a typical radial network of filaments, desmin does not. Instead it assembles primarily into large cytoplasmic aggregates with rudimentary filaments or short rod-like structures that extend out from these aggregates. Although other workers (Sjoberg et al., 1999; Dalakas et al., 2000) have recently indicated that desmin assembles into a filamentous network in SW-13 cells, their fluorescence microscopy images in fact show aggregated desmin filament organization that is very similar to that described herein, i.e. not in the form of an extended network. The difference between desmin and the other vimentin-like IF proteins does not appear to be related to an inability to assemble into filaments per se, but rather to the manner in which assembled filaments are organized within the cell.

The capacity of desmin to form extended IFs in SW-13 cells could be rescued by distinct mechanisms. Coassembly, either with vimentin or paranemin, allowed extended IFs. Since vimentin forms extended IFs on its own in these cells, this was an expected result. However, as paranemin is not capable of assembling into IFs on its own (Hemken et al., 1997), it appears that coassembly of desmin and paranemin results in filaments that have organizational properties that neither protein alone can provide. Since desmin and paranemin are normally coexpressed in muscle, it is likely that this has physiological significance and is consistent with the hypothesis that the large IF proteins provide some required interactions with other structures.

The lack of any observable effect of nestin or synemin is more difficult to interpret. In agreement with previous reports (Marvin et al., 1998; Bellin et al., 1999), neither of these IF proteins could form filaments in the absence of vimentin or desmin. Both of these large IF-proteins codistributed with desmin, indicating that both associated with desmin. The large C-terminal extensions of these proteins are believed to extend out from IFs and are likely candidates for interacting with other structures (Herrmann and Aebi, 2000), yet they only share about a 20% direct sequence identity (Bellin et al., 1999). The C-terminal synemin tail interacts with muscle α-actinin in vitro (Bellin et al., 1999), but it is not yet known if this is a property that is specific to synemin or shared with nestin or paranemin. Since SW-13 cells are non-muscle cells, it is possible that nestin or synemin require a muscle specific protein(s), not present in these cells, to mediate an effect on IF organization. The difference in the effect of paranemin from synemin or nestin on desmin IF organization in SW-13 cells, however, does suggest that these large IF proteins exhibit differences in the specificity of their interactions with other cell components.

The presence of a K8/K18 keratin network also appeared to allow extended desmin IFs. Although keratins and desmin do not form copolymers (Hatzfeld and Franke, 1985) but assemble into distinct IFs (Lu et al., 1993), there was clearly some codistribution of the two networks. The apparent dominant effect of desmin overexpression on the keratin network suggests that there is an interaction between the two IF networks and is in agreement with the observation of Yu et al., who reported that expression of truncated desmin in cells that contained keratin and vimentin IFs resulted in collapse of both endogenous IF networks (Yu et al., 1994). A similar interaction between keratin and vimentin networks was suggested on the basis of antibody microinjection experiments (Klymkowsky, 1982). These results indicate that, in addition to copolymer formation, desmin IFs can form an extended filament network that is dependent on an interaction with a separate network of IFs. Because keratins are not normal components of muscle it is unclear whether this interaction could have physiological significance. However, it has been reported that some extrafollicular reticulum cells of lymphoid tissues express both keratin and desmin (Franke and Moll, 1987; Gould et al., 1995), and there are numerous reports of tumor cells that coexpress keratin and desmin (for examples, see Garcia-Prats et al., 1998; Gerald et al., 1998; Hurlimann, 1994).

The difference in the capacity of desmin and vimentin homopolymers to form extended filament networks in these cells does not appear to be a property conferred by the non-helical NH₂-terminal head domains. This is in contrast to the situation in Xenopus, where domain-swapping experiments have shown that the NH₂-terminal head domains specified the organizational behavior of these proteins in muscle. When head-domain-swapped chimeras of desmin and vimentin were expressed in IF-free SW-13 cells, the organization of the assembled chimeric IF protein retained the characteristics of the body of the filament protein. A simple comparison of the aligned peptide sequences of mouse and Xenopus desmin reveals that the head domains are the most divergent regions of the polypeptides (not shown). These differences, together with the differences in cellular context, could be responsible for the differences in behavior that we observe. On the other hand, comparison of the rod and tail domains of human vimentin-like IF proteins (Quinlan et al., 1994) does not provide an immediate explanation for the unique organizational behavior of desmin compared with vimentin, GFAP or peripherin. There is some evidence that the C-terminal tail domain of vimentin interacts with actin based structures (Cary et al., 1994). However, relatively little is known about the role of specific IF structural domains in mediating interactions with other cytoplasmic structures, or how differences in the sequence of the other vimentin-like proteins might affect these interactions. This rather striking difference in the in vivo organization of desmin compared with the other vimentin-like proteins indicates that the sequence diversity of IF proteins reflects some cell-type specificity in the repertoire of interactions that individual IF proteins have evolved to provide.

The importance of heteropolymer formation in the assembly of keratin and neurofilament-type IFs is clearly established. Our studies extend this result to desmin IFs. While desmin can form homopolymeric filaments, the coexpression of paranemin and desmin appears to be required to generate an extended filament network. A variety of congenital human myopathies have been characterized by an abnormal accumulation of desmin (reviewed in Laing, 1999). Recent studies have shown that some of these desminopathies are associated with desmin mutations (Goldfarb et al., 1998; Munoz-Marmol et al., 1998; Li et al., 1999; Sjoberg et al., 1999; Dalakas et al., 2000) or mutations in α-crystallin, a desmin-associated chaperone (Vicart et al., 1998; Perng et al., 1999; Bova et al., 1999). Based on our data we would predict that desminopathies exist that involve mutations in paranemin, synemin and other, as yet unidentified genes (see Melberg et al., 1999; Vicart et al., 1996).

We are grateful to Martha Marvin for the nestin expression constructs, Katrina Trevor for the K8/K18 expressing SW-13 cells, and Ron Liem for the GFAP cDNA and the SW-13 cells that express peripherin. We would like to thank Alan Jones for the preparation of the thin sections for electron microscopy, and Stephanie Seiler for careful reading and suggestions for the manuscript. This work was supported by NIH grant HL51850 (R.M.E.), USDA-NRICGP Award 99-35206-8676 (R.M.R.), and the support of the NSF (M.W.K.), Muscular Dystrophy Association (M.W.K., R.M.R.), and the Colorado Chapter (M.W.K.) and Heartland Affiliate (R.M.R., R.M.B.) of the American Heart Association.